Free 30-minute AI & automation audit with a senior engineer. You keep the notes. Book yours

Use case · Travel & hospitality

AI that pre-fills booking support tickets and reviews every pull request

An LLM ticket assistant that extracts booking references and enriches helpdesk tickets with reservation data, plus an AI code reviewer that posts to Slack.

Client details are kept confidential. This use case describes the problem and what we built.

The brief

The client operates a B2B hotel-distribution platform with a busy support desk handling booking queries from travel buyers and hotels.

MultiGenesys built two practical AI tools around the team’s daily work: an assistant that turns free-text support tickets into structured, reservation-enriched records, and an AI code reviewer that gives engineers fast feedback on every pull request.

The challenge

What needed solving

Agents re-keying booking details

Support tickets about hotel bookings arrive as free text. Agents had to find the confirmation codes, look up each reservation in the admin system and copy hotel, room and date details into the ticket by hand.

Inconsistent ticket categories

Request type and category were set manually, so routing and reporting depended on how each agent read the message.

Review load on senior engineers

Every pull request needed a careful first pass. Senior engineers spent time on routine review comments instead of design and risk.

The solution

What we built

  1. Event-driven ticket pipeline

    New helpdesk tickets are queued on Amazon SQS. A Spring Boot consumer picks them up with a managed thread pool, so bursts of tickets do not slow the helpdesk.

  2. LLM extraction

    An OpenAI GPT model reads the ticket subject and body and returns structured output: booking and confirmation codes, request type and category.

  3. Reservation enrichment

    The service looks up each code in the booking platform and writes hotel, room, stay dates, refundability, cancellation cut-off and guest details back into the ticket's custom fields, then moves it into the open queue.

  4. AI pull-request reviewer

    An AWS Lambda function receives repository webhooks, fetches the diff, asks an LLM for review feedback and posts it to Slack. Developers rate the feedback in Slack, and ratings are stored in DynamoDB and S3.

  5. Safe rollout controls

    Feature toggles, CloudWatch metrics and token counting keep the assistant observable and let the team switch behavior on or off without a redeploy.

The outcome

What changed

Tickets arrive ready to work

Agents open tickets that already carry the reservation details they need, instead of starting with a lookup.

Consistent classification

Request type and category are set the same way every time, which makes routing and reporting more reliable.

Faster first-pass reviews

Every pull request gets automated feedback in the team's Slack channel, with developer ratings to tune the reviewer over time.

Technology used

  • Java
  • Spring Boot
  • OpenAI GPT
  • DeepSeek
  • AWS Lambda
  • Amazon SQS
  • Amazon DynamoDB
  • Amazon S3
  • Amazon CloudWatch
  • Slack API
  • Helpdesk API

Have a similar problem to solve?

Share a few lines about your project. A senior engineer — not a salesperson — replies within one business day.

Free AI audit Let’s talk